Vascular and parenchymal amyloid pathology in an Alzheimer disease knock-in mouse model: interplay with cerebral blood flow

We replaced exons 16 and 17 of the mouse APP gene with mutant sequences harboring the Swedish (K670N, M671L), Dutch (E693Q, amino acid 22 of Aβ), and London (V717I, all in APP770 numbering) FAD mutations together with humanized Aβ sequence by homologous recombination (Figure 1A and B), so that expression of the mutant APP and human Aβ is subject to its endogenous physiological regulations. In the knock-in vector, a pair of loxP sites were inserted to flank the truncated exon16 (that stops before the mouse Aβ sequence) and the neo-cassette; followed by the mutant exon 16 encoding the Swedish mutation and humanized Aβ sequence, and the mutant exon 17 containing the Dutch and London mutations (Figure 1A). Mice containing the targeted allele were then crossed with a germline Cre-deleter [22] to remove the truncated exon 16 and the neo-cassette and to express APP with the Dutch/Swedish/London mutations and humanized Aβ sequence.

Preliminary analysis showed that, similar to the APP Swedish/London knock-in mice, the APP Dutch/Swedish/London mice develop minimal Aβ deposits in their life time, we accordingly bred both lines with the PS1 M146V knock-in mice [23‐25] to facilitate the development of amyloid pathology within the aging time span of the mice. All subsequent studies were performed in the APP and PS1 M146V double homozygous background, listing the line with Swedish/London mutations and Dutch/Swedish/London as APP SL and APP DSL, respectively.

As expected, the APP DSL knock-in mice express similar levels of APP in comparison with that of wild-type mice and with an existing APP knock-in line expressing Swedish and London mutations and humanized Aβ sequence (APP SL) [20, 21] (Figure 1C and quantified in 1D). Therefore, the APP DSL and APP SL knock-in lines are otherwise identical except the addition of the Dutch mutation in the APP DSL mice. .

We performed sandwich ELISA to measure Aβ1-40 (Aβ40) and Aβ1-42 (Aβ42) concentrations in detergent soluble and insoluble fractions of brain lysates from young (Y, ~6 months), middle-age (M, 10 ~ 12 months), and old (O, 18 ~ 20 months) APP SL and APP DSL mice (Figure 2). Lysates from wild-type mice were used as a background control. We found that levels of Aβ40 and Aβ42 peptides in detergent-soluble fractions were comparable in young (light grey) and middle-age (dark grey) groups of APP DSL mice, but both were significantly elevated at old age (black bars Figure 2A and B). In contrast, a substantial elevation of insoluble Aβ40 and Aβ42 could already be detected in middle-aged APP DSL group compared to the young mice (Figure 2C and D), which is correlated with the beginning of amyloid deposition at this time frame (Figure 2E). Levels of insoluble Aβ, in particular Aβ42 (Figure 2D), increased sharply in old APP DSL mice, consistent with extensive amyloid deposition at this age (Figure 2E, 2F and quantified in 2G). These results indicate that the middle-age marks a critical stage when the APP DSL mice are no longer able to effectively maintain Aβ homeostasis, leading to elevated insoluble Aβ accumulation. The rise of soluble Aβ in addition to insoluble Aβ in old age likely reflects the further deterioration of the homeostasis. We did not isolate hippocampus for ELISA and plaque pathology due to the fact the plaque pathology was prominent only in the old age group of APP DSL mice in hippocampus, thus Aβ in total tissue lysate in ELISA and plaques in whole cortical area were used to indicate the progression of Aβ accumulation. In contrast to the APP DSL mice, Aβ levels and positive Aβ staining could only be detected in old APP SL mice (Figure 2, SL), but not in young and middle-age groups, when SL samples were prepared the same way as DSL samples. Therefore, introduction of the Dutch mutation leads to elevated steady-state Aβ levels across all ages and accelerated amyloid pathology.

We previously reported that the APP SL knock-in mice exhibit anxiety phenotypes starting at young age [20]. Analysis of young APP DSL mice revealed that these mice also display enhanced anxiety-like behaviors assayed by elevated plus maze with less distance traveled on the open arm than that of wild type mice (Figure 3A), and by conditioned fear assay with higher percentage of freezing time that was induced in a new environment (Figure 3B). These phenotypes persisted in middle-age animals. The enhanced freezing made it difficult to evaluate memory functions using the classical conditioned fear paradigm. We thus used the Morris water maze (MWM) to assess the spatial learning and memory abilities in APP DSL mice in young and middle-age groups and compared that with age-matched APP SL knock-in lines. APP DSL mice displayed normal spatial learning and memory at young age (Figure 3C and 3D). However, significant learning and memory deficits can be detected when APP DSL mice reached to middle age, as shown by longer latency to reach the hidden platform in the learning sessions (Figure 3E), and less duration in the targeted quadrant (Q3) during the memory probe test (Figure 3F). We subsequently investigated whether spatial related working memory is also affected in the APP DSL knock-in mice. Additional training was provided to allow the APP DSL reaching to the same level as the wild-type controls (Day 8) before they underwent reverse learning sessions (Day 9 ~ 12 in Figure 3G & H). During the following reverse-learning tests when the location of the hidden platform was switched from Quadrant 3 to Quadrant 4, the APP DSL mice extinguished the memory of the old location similarly as their wild type controls. However they formed the memory of the new location significantly slower than their wild-type controls (Figure 3H). The APP DSL mice were able to reach to a comparable level as wild type controls in the memory test at the last day of training, suggesting that their working memory deficits were at an early stage and could be corrected by intensive training (Day 12 in Figure 3H). Different from the APP DSL mice, the APP SL mice of the same age group demonstrated normal latency in locating the hidden platform in the training sessions as well as similar preference to the targeted quadrant in the probing test (Figure 3I and J). Old group of APP DSL mice exhibited the same anxiety behavior and learning and memory deficits as the middle age group, comparing with age and gender matched wild type controls (>18-month of age, n = 11 per genotype, Additional file 1). Since insoluble, but not soluble, Aβ were significantly increased in middle-aged APP DSL mice, the behavioral results indicate that the cognitive deficits are associated with insoluble Aβ accumulation. However, since ELISA was carried out using the whole hemisphere of the brain, we cannot exclude the possibility that there might be local changes of soluble Aβ levels in the middle aged group that were not observed from total tissue lysate.

Since the Dutch mutation is known to cause CAA in humans, we next examined the CAA pathologies in APP DSL mice and compared these with the APP SL mice. Minimal CAA could be detected in APP DSL animals at young age (Figure 4A). However, prominent CAA could be observed when the mice reached to middle-age and became more severe as the mice aged (Figure 4A and quantified in 4B). No CAA could be detected in APP SL mice until they reached to terminal age (>28 month, Additional file 2). Therefore, inclusion of the Dutch mutation leads to a simultaneous increase of both vascular and parenchymal Aβ pathologies starting at middle-age.

Since Aβ accumulation surrounding blood vessels is known to increase the risk of intracranial hemorrhage and hemorrhagic stroke in human brains [26‐28], we wanted to know whether similar pathology could also develop in APP DSL mice. We thus performed Prussian blue reaction to reveal iron released by hemosiderin in coronal sections of APP DSL mice and their age- and sex-matched wild-type controls and APP SL mice. No Prussian blue-positive signals could be spotted in cortical areas of wild-type or APP SL mice at any age (Figure 4C). While this was also the case in young and middle-age APP DSL mice, microhemorrhage became obvious in old APP DSL mice (Figure 4C, with average number of spots = 1.35 ± 0.46 per section, counted in 5 animals, 10 sections per animal). The microhemorrhages were almost exclusively located in the cortical region suggesting their association with amyloid buildup around cortical blood vessels. The fact that the onset of microhemorrhage is several months after the CAA supports the notion that CAA leads to the development of intracranial hemorrhage. Overall, the APP DSL knock-in mice recapitulate many of the pathological hallmarks associated with AD including age-dependent CAA and CAA-associated microhemorrhage.Because a prominent age-associated vascular Aβ pathology could be observed in the APP DSL mice, we asked whether CAA lead to changes in CBF in these mice. Using a MRI arterial spin labeling method, we measured relative cerebral blood flow (rCBF) in cortical regions of young, middle-age, and old wild-type and APP DSL mice (Figure 5). As expected, age-associated reduction of rCBF exists even in wild type mice (Figure 5B, open bars). Interestingly, the APP DSL mutant mice at young age already showed a mild reduction in rCBF compared to age-matched wild type controls, although this difference failed to reach statistical significance. As the mice aged, APP DSL mice develop much severe reduction in rCBF than their wild-type controls. We found that rCBF in APP DSL mutant was reduced to ~60% and ~40% of the wild-type controls at middle-age and old groups, respectively (Figure 5B), correlating with the severity of the CAA phenotypes (see Figure 4).

Having observed an exacerbating effect of CAA on age-dependent reduction of rCBF in the APP DSL knock-in mice, we set out to test if manipulation of blood flow could in turn affect Aβ pathology. We performed transverse aortic constriction surgery (TAC) on the left common carotid artery on 6 month-old APP DSL mice, a procedure that causes a substantial decrease of CBF on the anterior and medial regions of left side and a corresponding increase of CBF on the contralateral side of the same areas with minimal effects on the posterior regions. Non-invasive carotid Doppler ultrasound recording [29, 30] confirmed the success of the surgery, with ratios of the peak flow velocity of the left carotid artery to the right side ranging from 1:6 ~ 1:12. We terminated the mice at 1.5 or 3 months after the TAC surgery when animals reached to 7.5 and 9 months old, respectively, to evaluate the state of CAA and Aβ deposits in the two hemispheres. We did not detect hemorrhages in these brains, indicating that there was likely no acute ischemic insult as might be expected with complete carotid artery occlusion. No appreciable Aβ staining can be detected in mice 1.5 months post-TAC, suggesting that the surgical procedure or attendant anesthesia did not significantly accelerate the onset of the cerebral or vascular Aβ pathology. However, when analyzed 3 months post-TAC surgery, the degree of both CAA (Figure 6A and quantified in B) and parenchymal Aβ pathologies (Figure 6C and quantified in D) were much higher in TAC-treated mice compared to sham-treated controls, the latter were indistinguishable to age-matched untreated mice. Interestingly and unexpectedly, increases in CAA and plaque deposits can be detected in the entire brain irrespective of left-right hemispheres or anterior-posterior regions. With the exception of CAA severity between the anterior left and right hemispheres, there were no statistically significant differences in all other comparisons (Figure 6B and D). Therefore, perturbation of CBF has a global effect on CAA and Aβ plaque pathology. This could be due to that either circulating amyloid-aggregation-prone factors were produced by regional changes of CBF or that regional manipulation of CBF results in global impairment of Aβ clearance and thus increased overall amyloid deposits.In order to quantify soluble and insoluble Aβ40 & 42 concentrations in different cortical areas in TAC animals 3 months post operation, we dissected cortical tissue into left/right anterior/posterior parts from Zn-fixed sucrose cryo-protected frozen TAC and control brains for ELISA quantification (Figure 6E-H). Since Zn-fixation and sucrose protect increased tissue weight, we didn’t convert the final readout into per gram of wet tissue as we did in Figure 2A-D. Instead, detergent soluble fractions were normalized to the samples’ protein concentrations (2.5 mg/ml), and insoluble fractions were prepared from re-solubilized pellets of guanidine solution from 10 mg/ml total tissue lysates (details in Materials and Methods section). We found that insoluble Aβ42 is increased in all regions in TAC animals, correlated with plaque and CAA load analysis, changes in the soluble Aβ40 and 42 are more prominent in the anterior region than the posterior region, corresponding to the blood flow changes in the anterior area of the brain. Again, no differences between left vs right were observed, suggesting aberrant blood flow in either directions are deleterious.

All research and animal care procedures were approved by the Baylor College of Medicine Institutional Animal Care and Use Committee. PS1 M146V knock-in mice [23, 24] and APP knock-in mice carrying Swedish and London mutations and humanized Aβ region (APP SL) were previously published. APP DSL knock-in mice with Dutch/Swedish/London mutations and humanized Aβ region was generated in the similar manner as previously published APP/hAβ/mutC [25], except that the Dutch mutation (E618Q, APP695 numbering) was incorporated by site-directed mutagenesis when constructing the gene-targeting vector. Briefly, exons 16 and 17 of mouse APP were replaced by the mutant construct from the knock-in vector by homologous recombination, in which a pair of loxP sites flanked truncated exon 16 and neo-cassette, followed by mutant exon 16 with Swedish mutation (KM595-596NL) and humanized Aβ sequence (G601R, F606Y, and R609H), and then mutant exon 17 carrying Dutch mutation (E618Q) and London mutation (V642I) (Figure 1A). Genotyping was done by PCR using primers:

HL033, 5′-GTAATGCCTGTGTGGCCAAACACATG-3′ and

HL037, 5′-AAGTAATGGATTTGTTCTCCCAGGTCG-3′, which amplifies the loxP site. Expected PCR product from the wild type allele is 230 bp, and from the targeted allele is 270 bp.

The mice were then crossed with a germline Cre-deleter [22] to remove the truncated exon 16 and the neo-cassette and to produce APP with Dutch/Swedish/London mutations and humanized Aβ sequence. Crossing with PS1 M146V knock-in mice generate APP/PS1 homozygous double knock-in mice.

For Figure 2A-D, brain halves were weighted and homogenized in 4× (w/v) homogenization buffer composed of PBS, 1% Triton X100 and complete protease inhibitor cocktail (Roche). Total brain lysates were centrifuged at 100,000x g at 4°C for 1 hour and supernatants were diluted in half to be applied to Aβ1-40 and Aβ1-42 ELISA kit (Invitrogen, #KHB3481 and #KHB3544) following the company’s protocol. The resulting pellets (insoluble fractions) were resuspended in 5 M guanidine HCl/50 mM Tris HCl, pH8.0, rotated at room temperature for 4 hours, and centrifuged at 20,000×g for 20 minutes. Supernatants from the last step centrifugation were diluted 500 ~ 2000 fold before applying to the ELISA kits. Each sample was analyzed in duplicate. Three to five male animals were used for each genotype at each age group.For Figure 6E-H, 9-month-old male mice (with or without transverse aortic constriction surgery (TAC) at 6 months) were perfused and fixed using Zn-fixatives. The retrieved brains were cryoprotected by soaking in 30% sucrose for 36 hours and frozen in −80°C before ELISA experiment. To obtain cortical tissue for left/right anterior/posterior region, mouse brain was put into acrylic 1 mm adult mouse brain matrix for coronal slices to remove olfactory bulbs and cerebella. Cutting in the mid-line separates left vs right half of the brain, cutting in the 5 mm slot separates anterior and posterior region. Corresponding parts of cortex were then peeled off and detergent soluble and insoluble lysate were prepared as mentioned in the above paragraph. Due to Zn-fixation and sucrose cryoprotection, the tissue weighted heavier than un-perfused tissue used for Figure 2A-D, rendered it impossible to deduce comparable Aβ concentration in per gram of wet tissue as before. Thus, ELISA results for soluble samples were normalized to the protein concentration of 2.5 mg/ml. For insoluble fractions, 300 ul of 10 mg/ml of total tissue lysate were centrifuged for 1 hour at 100,000×g 4°C; resulting pellets were re-solubilized in 200 ul of guanidine solution as in the above paragraph, and diluted 20× before applying to ELISA plates. All data in this group were presented as pg/ml instead of converting to pg/g of wet tissue. All samples were analyzed in duplicates on ELISA plates, and 3 male animals were used for each group of treatment.

Antibody staining on paraffin-embedded brain sections was performed as described [24]. In particular, paraformaldehyde (PFA) perfused brains were cut into 10 μm paraffin sections. The sections were deparaffinized in xylene, then rinsed with series of ethanol and water mixture. Antigen-retrieval was done by incubating sections with 70% formic acid for 6 minutes. Endogenous peroxidase activity was quenched by incubating the slides in 0.3% H₂O₂ for 30 minutes. The slides were rinsed with Tris-buffered saline (TBS; 50 mM Tris-Cl, pH 7.5 and 250 mM NaCl). Nonspecific epitopes were blocked for 30 minutes with 3% normal goat serum, 0.4% Triton X-100 in TBS. Primary antibodies against smooth muscle actin (Sigma, A2547) and/or Aβ (Cell Signaling, #2454) were diluted 1:1000 in blocking buffer, and incubated overnight at 4 ° C in a humid chamber. Sections were then washed 3 x 5 minutes each in TBS and incubated with anti-mouse-Alexa 488 and/or anti-rabbit-Alexa 555 secondary antibodies at room temperature for 1 ~ 2 hours. Pictures were taken with a Leica TCS SPE microscope and the images were processed with ImageJ (National Institute of Health). Mice underwent aortic banding surgery were fixed in a similar manner but using Zinc-fixation [51] (0.1 M Tris–HCl pH7.4, 0.05% Calcium acetate, 0.5% Zinc acetate, 0.5% Zinc chloride) instead of PFA. Immunohistochemistry on Zn-fixed brain sections were done using alkaline-phosphatase standard method (Vector Laboratories, AK-5000) and corresponding alkaline phosphatase substrate kit (Vector Laboratories, SK-5100).

Cryosections of WT, SL, and APP DSL mice were mounted on glass slides and dried overnight. Slides were immersed in freshly mixed 4% Potassium Ferrocyanide with 4% of hydrochloric acid, and microwaved for 30 seconds. Then rinsed in tap water for 3×10 minutes. Next, slides were counter stained using nuclear fast red at room temperature for 5 minutes, rinsed under running tap water for 1 minutes, dehydrated quickly in 95% and 100% ethanol, cleared in xylene, covered by mounting medium and coverslips for imaging. 3 ~ 4 male animals were used for each genotype at old age group as well as APP DSL in middle age group.

Mice were housed in a 12 hour light/12 hour dark cycle. All behavior tests were done on naïve male mice at either young (4 ~ 6 month old) or middle (12 ~ 13 month old) ages. 10 ~ 18 male animals were used for each genotype at each age group. Mice were tested in elevated plus maze first, conditioned fear experiment two days after, and the Morris water maze one week later.

The testing room was at 700 LUX illumination, with 60 dB white noise. Mice were brought to the room in their home cages and allowed to acclimate for 1 hour before starting the test. The testing apparatus is a plus-shaped platform made of white plastics, elevated 50 cm above the ground, with each arm 30×5 cm. Extended from a 5×5 cm center area are two opposite high-walled enclosed arms and two open arms (with a shallow rim along the side). The mouse being tested was placed in the center area and allowed to explore the plus maze for 10 minutes. Entrances, duration, and distance traveled in each region were recorded by a camera connected to an automated ANY-maze tracking system (Version 4.50; Stoelting, Co., Wood Dale, IL, USA). The platform was cleaned with 70% isopropanol after each trial. All mice were novel to the test, and EPM is the first behavior test performed on these mice.

Mice were transported to the holding room in their home cages 30 minutes prior to the test on the training and test days. On the first day (training sessions), each mouse was transferred to the testing room, placed in the isopropanol-cleaned chamber for 2 minutes, and then received a mild foot shock (2 seconds, 0.7 mA) delivered from the grid floor each time after hearing a 30-second tone (conditioned stimulus); 2 minutes later, the tone and the shock would repeat again, and the mouse would be removed from the chamber and transferred back to the holding room. 24 hours after the training session, mice were first tested for freezing behaviors in the same chamber without the tone stimulus (identical context). 2 hours later, mice were tested for freezing behavior in an altered environment. Each mouse was transported by the same experimenter to the same experiment room, but into the 70% ethanol-cleaned chamber with vanilla smell (different odor cue), triangle space (different testing space), and smooth white plastic floor (different tactile cue), for a total of 6 minutes. The first 3 minutes were quiet and the last 3 minutes the tone was played. Movements of the mouse were recorded by a video camera and analyzed by FreezeFrame software (Version 2; San Diego Instruments).

Morris water maze tests were performed as previously published [52‐54]. Briefly, mice were trained to locate a square platform (9 × 9 cm) hidden in quadrant 3 of a circular water pool within 60 seconds. In the learning sessions, each mouse was trained in total of 32 trials for four consecutive days. Each day, each mouse received 2 blocks of training, and in each block, the mouse was trained for 4 successive trials. At the end of the 32 trials, the mouse was subjected to a probe test during which the platform was removed and each mouse would swim for 60 seconds to search for the platform. APP DSL/PS1 double knock-in mice, together with their wild-type control group, received a simplified training of 8 trials on Day 5, in which mice were returned to its holding cage after each swimming trial. The mice underwent reverse learning sessions the following week, in which the location of the hidden platform was moved from quadrant 3 to quadrant 4. Mice were trained in 4 trials each day for another four consecutive days. On Day 9, before the reverse learning session, mice were exposed to a probe test of 60 seconds, in which the platform was removed for the 60-second probing period but placed back at its old location at the end of the session (to minimize the extinction effect of probe run). Probe tests on Day 10 ~ 12 were done after the training sessions, and the hidden platform was put back into the pool at the new location at the end of each test. For all the swimming sessions, the training/testing mouse was rested on the platform for 10 seconds before it was submitted to the next trial or transferred back to its holding cage. The time for mice to locate the hidden platform during learning and reverse learning sessions (escape latency), as well as the time they spent in each quadrant during the probe tests (quadrant duration) were recorded by Ethovision tracking system (Noldus Information Technologies).

All mice were initially anesthetized with 5% isoflurane in oxygen and then transferred to an animal holder within the magnet where they were maintained on 1.5-2% isoflurane in oxygen during imaging. Respiration rate was monitored using a respiration pad (Small Animal Instruments) and maintained at approximately 30–35 breaths per minute to minimize motion artifacts during imaging. Body temperature was monitored with a rectal temperature probe and was maintained at a constant 37°C using an air heating system (Small Animal Instruments) throughout the relative cerebral blood flow (rCBF) measurement. MRI imaging was carried out utilizing a 9.4 T, Bruker Avance Biospec Spectrometer with a 21 cm horizontal bore and a 35 mm volume resonator (Bruker BioSpin).

A slice of interest within the brain was selected using setup scans aligned to where the rostral segment of the hippocampus first appears. The rCBF was measured using an arterial spin labeling (ASL) MRI imaging protocol. A flow alternating inversion recovery-echo planar imaging protocol with the following parameters was utilized: TR = 7555.38; TE = 16.73; NA = 1; matrix = 64×64; field of view = 1.5cmx1.5 cm; slice thickness = 1.5 mm. Both a selective and nonselective ASL image was acquired and the absolute blood flow was then calculated within two regions of interest within the cortex using the formula rCBF = λ(1/T1selective – 1/T1global) where λ is the blood brain partition coefficient which in the mouse brain is equivalent to 0.9. Five male mice per genotype of young and old groups, and 8 per genotype of middle aged group were tested in this experiment.

Transverse aortic constriction and Doppler evaluation of the blood flow were performed in mice using the method described in previous publications [55]. In brief, to produce a transverse aortic constriction, mice were anesthetized with an initial dose of 5% Isoflurane and maintained on 2% isoflurane via inhalation, and were placed on a heating pad at 38.6°C to maintain body temperature during the surgical procedure. The animal was placed supine and endotracheal intubation was performed. The endotracheal tube was connected to a Harvard volume-cycled rodent ventilator cycling at ≈ 150 breaths/minutes and with sufficient volume for adequate expansion of the lungs, with a tidal volume of approximately 0.20 -0.25 ml. Oxygen (100%) and maintenance dose of 1–1.5% isoflurane was supplied to the inflow of the ventilator during operation. Under a microscope, the chest cavity was opened and the aortic arch was exposed after deflection of the thymus. After the transverse aorta was isolated, a 7–0 nylon suture was placed around the aorta between the origins of the innominate and left carotid arteries. A 27-gauge (0.4 mm diameter) needle was placed against and tied to the aorta with the suture. Then the needle was removed promptly to yield a constriction with a diameter approximately equal to that of the needle, and produce a 60-80% constriction. In sham-operated mice, the suture was placed around the aorta but not tightened. The chest wall was closed and the mice were allowed to recover from anesthesia under oxygen and a heat lamp to maintain body temperature. Doppler velocity was measured in the right and left carotid arteries to confirm that the blood flow velocity was reduced on the left and increased on the right side and the severity of the constriction is assessed by the ratio of the left to right peak carotid flow velocities [55]. Brains from 4 male mice per treatment group were used for paraffin section/immunohistochemistry analysis, and 3 male mice per genotype for 3 month old group were used in Aβ40 and 42 ELISA analysis.

All raw data (from ImageJ, ELISA assay, and behavior tests) were tabulated in Microsoft Excel. Statistical analyses were performed using student t-test in Excel, or two-way ANOVA in GraphPad Prism4 (GraphPad Software, http://​www.​graphpad.​com). Data were presented as average ± SEM (standard error of mean). NS stands for no significance; *, p < 0.05; **, p < 0.01; and ***, p < 0.001. For Morris Water Maze assays, escape latency and distance traveled data were analyzed using a two-way (genotype × trial block) ANOVA with repeated measures. Search data from the probe trial were analyzed by individual one-way ANOVA and Newman-Keuls post comparison tests. One-way ANOVA was used to compare the search time and platform crossing data for the training quadrant only between wild-type and mutant mice.